spi in dev-esp32-idf4 seems broken

[Philip Gladstone]

Expected behavior

device:transfer({address=1, rxlen=1, txdata=string.char(0)})

ought to work

Actual behavior

It actually says

spi_master: check_trans_valid(687): rx length > tx length in full duplex mode

Root cause

It appears that the code in spi_master.c is broken:

in lspi_device_transfer the following code is missing the line that I've marked

    trans.cmd  = opt_checkint( L, "cmd", 0 );
    trans.addr = opt_checkint( L, "addr", 0 );
    //
    rx_len = opt_checkint( L, "rxlen", 0 );
    //
    trans.flags |= opt_checkbool( L, "addr_mode", false ) ? SPI_TRANS_MODE_DIOQIO_ADDR : 0;
    //
    lua_getfield( L, stack, "mode" );
    trans.flags |= options_flags[ luaL_checkoption( L, -1, options[0], options ) ];
    lua_pop(L, 1);       //     MISSING LINE  ---------------------------------------------------
    //
    data_len = 0;
    data = opt_checklstring( L, "txdata", "", &data_len );

Without this line, the txdata doesn't get fetched from the right place (as the lua_getfield pushes an element onto the stack).

This leads to a more general question -- does anybody have working spi code? I'm trying to talk to a device but it doesn't appear to work. I may have to break out the logic analyzer....

[Philip Gladstone]

Also, if you want to use the cs config item when you create the device, then you need to configure that as a gpio output pin and write the inactive state to it. I wonder if this should just be builtin to our device creation function. I don't think that it would do any harm -- and might save people a bunch of time. [I had to get my logic analyzer out to see what was going on!]